CommandCodeAI / CommandCodeAI/command-code
feat: `cmd skills update` -- record install source so skills can be updated
還沒有人認領這個 Issue。
- 主要語言
- 沒有語言資料
- 星號
- 4k
- 分支
- 350
- PR 合併指標
- 30 天內沒有已合併 PR
描述
Problem
cmd skills add does not record where a skill came from. Once installed, there is no manifest mapping skill name to source repo -- nothing in ~/.commandcode/ or the skill directory preserves the owner/repo[/path][@ref] spec used at install time.
Consequence: there is no way to update an installed skill from within the CLI. cmd skills only has add / remove / list (unchanged since v0.18.2). The only update path is:
- Rediscover the source repo yourself (out of band -- the CLI can't tell you)
- Re-run
cmd skills add <owner/repo/path> -f
For skills installed from multi-skill repos or subpaths (e.g. cmd skills add acme/repo/path/to/skill), rediscovering the exact spec is genuinely error-prone.
Prior art
- Mods already solve this in Command Code itself:
cmd mods updatereinstalls and reconciles every configured mod package, because mod sources are tracked insettings.json. Skills are the odd one out. - GitHub CLI shipped
gh skill(v2.90.0, April 2026), which records provenance metadata at install time and supportsgh skill update [name|--all]across agent hosts.
Proposal
- At
cmd skills addtime, write the resolved source spec (owner/repo[/path][@ref]plus installed commit SHA) to a manifest -- either askills.jsonalongside the skills directory or a per-skill metadata file. - Add
cmd skills update [skill-name]andcmd skills update --all: re-fetch from the recorded source and overwrite in place (same semantics asadd -f). cmd skills listcould then also show source + whether upstream has moved.
Part 1 is the load-bearing piece -- without provenance, no update verb is possible. Backfill for already-installed skills can be manual (prompt on first update, or accept a spec argument to re-pin).
Tested on v1.5.0 (macOS); docs at commandcode.ai/docs/skills confirm only add/list/remove exist.
貢獻指南
這個儲存庫沒有索引到貢獻指南
從這裡開始
- 先讀完整個 Issue,再讀專案的貢獻指南。
- 在 Issue 下留言說明你要接手 —— 這能避免兩個人做同樣的事。
- Fork 儲存庫,在一個分支上完成修改。
- 送出 Pull Request,並在描述裡引用這個 Issue 編號。
研究方向
從現有的 cmd skills add、list 和 remove 進入點開始,然後比較 cmd mods update 使用的 settings.json 中的來源處理方式。完成的標準是安裝能夠保留來源規格和 commit SHA,同時 cmd skills update [skill-name] 和 --all 可以重新擷取並覆寫已安裝的 skills;檢查 skills 文件以了解目前的命令介面。
由索引模型根據 Issue 內容生成。
評估
- 領域
- cli
- Issue 類型
- 功能
- 難度
- 4/5
- 預估耗時
- 3-5 天
- 活躍度
- 冷清
- 描述清晰度
- 基本清楚
- 新手友好度
- 48/100